Posted: Jun 19, 2012 6:50 pm
So it is difficult to figure out in what way they are fixed. We are talking here of salivary amylase gene copy number variation. Copy number variation has been shown to occur between monozygotic twins who presumably started out with identical genomes. In other words if they can't be shown to be fixed even in an individual how can they be shown to be fixed in an unobserved population?
[1] Somatic mosaicism is irrelevant to evolution. Germline mutations are. The fact that you get massive variations in regions of the genome in somatic cells does not actually impact how the mutation is fixed or inherited. If said variations are due to hotspots of duplication, those hotspots themselves will be subject to selection. To sum up, evolvability itself is potentially a selectable trait.
Also, since you mention unobserved populations, we use the general statistical tools of Confidence Intervals that facilitate extrapolations from observed populations.
[2] It is rather amusing to confuse the fixation of mutations/alleles with fixations of copy numbers themselves. See previous point for an explanation of why this is the case.
[3] New CNVs can occur in the presence of pre-existing ones just like new SNPs can form in the background of pre-existing ones, unless of course there is an initial loss of copy number to the point of eliminating the gene.
In the case of loci prone to massive copy number variations, that itself is selected for or prone to elimination. Just like SNPs, CNVs also have a distribution that ranges from being beneficial to being deleterious (generally tending towards being deleterious http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2577867/ ), you still don't get organisms developing CNVs by choice to meet particular needs to eventually meet their visions of some kind of archetype. While cellular stress does seem to increase the accumulation of CNV's, there is no arbitrary trend in favour of beneficial CNVs, for instance.
[4] I am not mistaking the duplication for a single change that happened so many years ago et cetera. CNVs and SNPs may have different mechanisms by which they come about (involvement of DSB repair mechanisms for CNVs as opposed to polymerase proofreading errors for SNPs) but the principle is still the same...in terms of fitness effects you get a spectrum subject to natural selection.